The aim of this proposed research is the development of a method for characterizing the speech patterns of individual talkers. This characterization will be in the form of procedure for the adjustment of the parameters of a speech synthesizer so that it generates output that sounds like a given talker. The procedure will consist of two steps: 1) automatic extraction of a set of acoustic descriptions from recorded utterances of the talker, and 2) adjustment of parameters and rules on a synthesis-by-rule system based on the extracted descriptions. Phase I of the project will examine the feasibility of this approach using a limited number of talkers and sentences. The product that will be developed will be a component of a speech synthesizer that is capable of automatically modifying the rules of the synthesizer so that it wounds like any chosen talker. The system will have important application for laryngectomy and other speech- impaired patients who want synthetic speech that is close to their normal voice. The system will also have applications in speech pathology since it can characterize deviant voices. The system will have applications in speech recognition and talker verification as well. PROPOSED COMMERCIAL APPLICATION: Potential commercial application not available.